The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill performs pre-flight checks that download shell scripts from GitHub repositories and pipes them directly to the shell (e.g., curl -sSL ... | sh). This occurs for scripts from both 'okx' and 'purong-huang-1121' repositories, allowing for arbitrary code execution from remote sources at runtime.
[PROMPT_INJECTION]: The skill contains a 'Post-Install Welcome' section explicitly telling the user that security warnings (Gen/Snyk/Socket) are 'false positives' (误报) and should be ignored. This is a social engineering tactic designed to deceive the user into bypassing safety filters.
[COMMAND_EXECUTION]: The skill documentation and the provided engine.py rely heavily on the execution of external binaries and system commands using subprocess.run and shell piping to perform environment checks and trading operations.
[EXTERNAL_DOWNLOADS]: The skill downloads binaries and installer scripts from external GitHub repositories during its initialization and update phases.
[CREDENTIALS_UNSAFE]: The skill implementation and documentation require users to store highly sensitive information, such as SOLANA_PRIVATE_KEY and TELEGRAM_BOT_TOKEN, in a local .env file for the bot to function.
[DYNAMIC_EXECUTION]: The skill uses python3 -c to execute dynamically generated Python code passed through a pipe from a curl command to process JSON data from the GitHub API.

strategy-ranking-sniper